You are on page 1of 36

Does an Asset Management Firms Stock Holding Made in Response to Buy-Side Analysts

Prior Recommendations Induce Subsequent Forecast Optimism?

Hun-Tong Tan*
Nanyang Technological University

Rong-Ruey Duh
National Taiwan University

Shean-Bii Chiu
National Taiwan University

Shu-Hsing Li
National Taiwan University

August 31, 2015

Corresponding author:

Hun-Tong Tan
Room S3-01C-78, Nanyang Business School
Nanyang Technological University
Tel: 65 67904819
Fax: 65 67937956
Email: ahttan@ntu.edu.sg

We appreciate comments from Wei Chen, Jun Han, Lukas Helikum, Terence Ng, Premila
Shankar, Seet Koh Tan, Elaine Wang, Yao Yu, and Bo Zhou, and research assistance by Lukas
Helikum, Li Xiao, Feng Yeo, and Yao Yu. Hun-Tong Tan acknowledges the generous funding of
the UOB Endowment.

Electronic copy available at: http://ssrn.com/abstract=2675686


Does an Asset Management Firms Stock Holding Made in Response to Buy-Side
Analysts Prior Recommendations Induce Subsequent Forecast Optimism?

Abstract

In this study, we examine whether buy-side analysts forecasts are biased when they cover stocks
that are held by the mutual fund. Specifically, we conduct an experiment with buy-side analysts
and fund managers where we manipulate whether the forecasts are released to colleagues and
clients or kept private, and whether the mutual fund has made an investment in the stock based
on the recommendation of the analyst. We posit and find evidence that when forecasts are made
available to other colleagues and clients, participants forecasts are more positive when the
mutual fund has a holding (versus no holding) in a stock based on the buy-side analysts
recommendations. These forecasts are also more optimistic than a control group where
participants have an accuracy goal. This effect disappears when the forecasts are kept private.
Our findings indicate that institutional features in the asset management setting create conditions
for buy-side analysts forecasts to be optimistic.

Keywords: buy-side analysts; fund managers; forecasts; optimism

Data Availability: Contact the authors.

Electronic copy available at: http://ssrn.com/abstract=2675686


INTRODUCTION

Buy-side analysts are important financial intermediaries who work for asset management

firms and make stock recommendations to fund managers in these firms. Their interpretation of

earnings- and share price-relevant information benefits the segment of investors who place their

funds with the asset management firms that these analysts work for. In this study, we conduct an

experiment to examine two factors that may influence buy-side analysts forecast optimism

whether the mutual fund has made an investment in the stock based on a previous

recommendation of the analyst, and whether the forecasts are released to colleagues and clients

or kept private.

The first factor is an institutional feature present in the buy-side analyst environment.

Buy-side analysts follow up and issue forecasts on stocks already held by the mutual fund, and

they also cover stocks not currently covered by the fund in sourcing for new buy ideas. Further,

in the buy-side analyst environment, a stock held by the mutual fund is also likely one that has

been recommended by the buy-side analyst (Frey and Herbst 2014). Psychology research on

motivated reasoning (e.g., Kunda 1990) and strategic justification (e.g., Caldwell and OReilly

1980; Leary and Kowalski 1990; Schoorman and Holahan 1996) suggests that this can lead to

optimistic forecast bias. Further, although these two aspects (holding, stock recommendation)

naturally co-vary in practice, we attempt to disentangle their separate effects in our study

knowing the cause of the judgment bias (if it exists) allows asset management firms to allocate

their effort accordingly to de-bias the buy-side analysts.

The second factor we examine, whether the forecasts are released to colleagues and

clients or kept private, is another institutional feature present in the buy-side analyst

environment, and one that likely magnifies the effect we discussed above. Buy-side analysts

3
forecasts and recommendations are issued with the express purpose of selling stock ideas to the

asset management firm and its clients (Groysberg, Healy, and Chapman 2008). 1 Portfolio

managers we interviewed indicated that, for marketing and client relationship purposes, these

buy-side analysts recommendations/forecasts are disseminated to portfolio managers and

selected clients of the asset management firm.2 These parties are important stakeholders to whom

buy-side analysts are accountable, with a preference for the stocks earnings prospects to be good

should they have a stake in the stock. In particular, if the asset management firm has a stock

holding (e.g., by the fund manager on behalf of a client) based on the buy-side analysts prior

recommendation, the biasing effect of this stock holding discussed earlier is likely heightened if

buy-side analysts are aware that their forecasts will be released to these parties. For instance,

fund managers and clients are likely to be unhappy should they receive pessimistic earnings

forecasts from the buy-side analyst after having acted on the latters previous recommendation to

buy a stock. Psychology research (e.g., Tetlock, Skitka, and Boettger 1989) suggests that under

these circumstances, buy-side analysts are motivated to bias their forecasts upwards to please

these parties.

Investigating these issues is important. The quantum of funds under the charge of asset

management firms is significant, with the value of professionally managed assets rising by 8

percent to US$56 trillion in 2010 (Boston Consulting Group 2010a, 2010b). Buy-side analysts

have significant influence on the asset allocation by asset management firms (Cheng, Liu, and

Qian 2006; Frey and Herbst 2014; Groysberg, Healy, and Serafeim 2013), and judgment biases

by buy-side analysts have economically significant financial consequences. Research involving

1
These recommendations (and the associated earnings forecasts) are likely buy recommendations because asset management
firms generally have restrictions on short-selling (Groysberg et al. 2008). For instance, Frey and Herbst (2014, Table 2) indicate
that only 14 percent of buy-side analysts recommendation revisions involve a downgrade to underperform or sell, and 73
percent of the revisions start from a hold or buy position.
2
The webpage by Graeme Pietersz, former equity analyst, makes a similar point about buy-side analysts reports being made
available to clients (http://moneyterms.co.uk/buy-side-analyst).

4
buy-side analysts is rare, largely because of data availability issues (see Willis 2001 and

Groysberg et al. 2008 for exceptions).

Of direct relevance to our study is an archival study by Willis (2001), who analyzes

mutual fund manager forecasts that were publicly released via the media.3 He finds that stock

holdings are positively associated with greater optimism bias. Specifically, he uses a subset of 60

mutual fund managers who issue more than one forecast for a particular day, and finds a positive

correlation (without control variables) between forecast optimism and dollar investment in the

stocks. However, as Willis (2001, p. 721) acknowledges, his study cannot disentangle whether

increased stock holding results in larger forecast optimism, or alternatively, greater forecast

optimism induces greater stock holding. Further, all the forecasts issued by fund managers in

Willis sample relate to stocks that the mutual fund holds, so it is not possible to determine

whether the presence versus absence of a stock holding leads to forecast optimism. The forecasts

in his dataset are publicly available, and it is not possible to disentangle the effect of holding

versus public availability of the forecasts. In short, absence of an appropriate comparison group

in the dataset precludes examination of the causal influence of these variables.

We conduct an experiment to examine our research questions. An experiment allows us

to provide direct evidence on the causal relationships among these variables, holding constant the

attributes of the company and the information environment. A constraint in the use of archival

analyses is that it is difficult to determine whether stock holding is the antecedent or

consequence of forecast optimism. Our use of an experimental approach avoids this issue.

Further, an experimental approach allows us to separate out the effects of stock holding from the

3
Willis (2001, p. 708) considers mutual fund managers to be a subset of buy-side analysts. Senior members of asset management
firms tell us that in some asset management firms, buy-side analysts have distinct career paths; in many others, aspiring fund
managers first serve as buy-side analysts before they progress to be fund managers. While fund managers make investment
decisions on behalf of their clients, they also routinely make stock recommendations/forecasts (see also Frey and Herbst 2014).

5
effects of stock recommendation, both of which naturally co-occur in practice and are therefore

impossible to disentangle using archival data.

In our experiment with experienced buy-side analysts/fund managers, we directly

manipulate whether the asset management firm has a holding in a stock based on the

recommendation of the buy-side analyst, and whether the participants assume that the forecasts

they make will be kept private, or made publicly available to other fund managers and clients.

Our dependent variable is the earnings forecasts made by the participants. Manipulating the asset

management firms stock holding addresses the difficulty experienced in archival research of

eliminating the alternative explanation that it is forecast optimism that leads to greater stock

holding by the firm, and not the reverse. By also separately manipulating the public/private

nature of the forecast release, we address the issue of whether the findings in Willis (2001) are

driven by the public nature of the forecasts in his database.

We find evidence that forecasts which are made available to other colleagues and clients

are more positive when the mutual fund has a holding (versus no holding) in a stock based on the

buy-side analysts recommendations. This effect disappears when the forecasts are kept private.

The forecasts by a control group where participants are instructed that their goal is to make

accurate forecasts are lower than those made by the holding/public group, and not different from

those in any other condition. This indicates that the holding/public condition leads to an upward

bias in analysts forecasts. In contrast, the fact that there is no apparent bias in the

holding/private condition suggests that the presence of a stock holding does not bias buy-side

analysts forecasts, at least when they make these forecasts privately.

As we mentioned earlier, a stock held by the mutual fund is also likely one that has been

recommended by the buy-side analyst (Frey and Herbst 2014). We disentangle the effects of

6
stock holdings per se versus stock holdings based on the buy-side analysts recommendations,

conditional on public dissemination of the forecasts. We do so by comparing our main results

with those from a modified public/holding condition where the participants are not told that the

stock holding is based on the buy-side analysts recommendations. Our results indicate that the

mere holding of a stock by the mutual fund is insufficient to create optimism in the forecasts

issued by buy-side analysts, at least in the public dissemination condition. However, given that it

is somewhat less common to have a fund hold a stock without the current analyst having made

an earlier recommendation, the threat of analyst forecast bias in a public release setting remains.

In a post-experiment within-subjects test, we find that participants are aware of the

negative impact of disseminating below-consensus forecasts publicly (versus privately). They are

also aware of the negative impact on client relationship of disseminating below-consensus

forecasts when there is a stock holding. We also conducted supplementary experiment using

buy-side analysts/fund managers in which the two dissemination conditions were manipulated

within-subjects in a setting where the mutual fund has a holding in a stock based on the buy-side

analysts recommendations. Results indicate that participants forecasts are higher in the public

condition than in the private condition, suggesting that the effect is conscioushad the effect

been unconscious, it should have disappeared once the public/private difference is made salient

in the within-subjects design. These results, along with our finding that, in the presence of a

stock holding, participants can switch from being objective in a private release setting to being

more biased in a public release setting, suggest that the biasing effect of stock holding in the

public condition is likely a conscious strategic bias.

Our study contributes to the literature by providing a theoretical account and empirical

evidence on the causal relation between an asset managements holding of a stock through a buy-

7
side analysts previous recommendation and buy-side analysts forecast optimism, and how this

effect is moderated by the dissemination of his/her forecast. Our results extend the archival

finding of a holding effect by Willis (2001) in three aspects. First, that the holding/public

condition (akin to Willis setting) but not the holding/private condition leads to an upward bias in

analysts forecasts suggests that the stock holding effect in Willis (2001) may be attributable to

the public nature of the forecasts in his sample. Second, at least in a public forecast release

setting, it is the stock holding given a prior stock recommendation and not stock holding per se

that is likely driving buy-side analysts forecast optimism. This also has a practical implication

for asset management firms in terms of taking steps to remedy the source of biased forecasts.

Finally, we provide evidence that our findings are the result of a conscious, not sub-conscious

mechanism.

The remainder of this paper is structured as follows. In the next section, we provide

theoretical background and develop our research hypothesis, which is followed by our research

design. The fourth section presents the results. The final section concludes our paper.

BACKGROUND AND HYPOTHESIS DEVELOPMENT

While studies involving buy-side analysts have been relatively scarce, there is extensive

research on sell-side analysts behavior. Sell-side analysts work for brokerage firms and issue

stock recommendations for the clientele of the brokerage firms (see Bradshaw 2011 for review).

These studies conclude that sell-side analysts issue optimistic forecasts because of incentives to

attract underwriting business (Lin, McNichols, and OBrien 2005), to generate trading

commission (Cowen, Groysberg, and Healy 2006), or to develop good relationship with

management (Ke and Yu 2006). McNichols and OBrien (1997) offer another explanation,

suggesting that sell-side analysts selectively report their forecasts and stock recommendations by

8
reporting those only when they hold favorable views. Some studies also suggest the possibility of

a cognitive bias (Affleck-Graves, Davis, and Mendenhall 1990).

Research on sell-side analysts do not necessarily generalize to buy-side analysts because

the institutional settings are different. For instance, unlike sell-side analysts, buy-side analysts do

not have incentives to please management of the firm they are analyzing. Further, the asset

management firms holding of a stock based on a buy-side analysts recommendation is unique

to the buy-side analyst setting.

Little is known about the institutional setting within which buy-side analysts operate, and

how these institutional features influence the forecasts they issue. Archival data on buy-side

analysts forecasts are generally unobservable as they are proprietary in nature. Hence, there

have been few studies on this issue, with two notable exceptions: Willis (2001) and Groysberg et

al. (2008) analyzed proprietary data of buy-side analysts from a mutual fund.4 Both conclude that

buy-side analysts forecasts are optimistically biased. However, neither examines specific

institutional features in the buy-side analyst environment that influence this forecast optimism.

Studies involving buy-side analysts as participants in experimental settings have also been

relatively scarce, and those that do so predominantly use them as proxies for sophisticated

investors (e.g., Hopkins, Houston, and Peters 2000; Miller and Sedor 2014), but not as direct

subjects of interest (one exception is a study by Ashton and Cianci 2007, which compares the

forecasts made by buy-side and sell-side analysts).

4
Groysberg et al.s (2008) dataset partially addresses a concern in Willis (2001) dataset that forecasts which are publicly
available tend to be optimistic to manage public expectations. Note that fully private forecasts do not exist in Groysberg et al.s
(2008) sample, just like those in Willis (2001) sample. Some of these forecasts are open to the firms fund managers and
potentially clients as well, who likely have taken positions based on these buy-side analysts recommendations (see Frey and
Herbst 2014). In addition, because the mutual fund analyzed in Groysberg et al. (2008) is one that prohibits short selling, it is
likely that the forecasts made, including those for newly covered firms, are those that the analysts recommend holding (or had, at
some point in the past, recommended holding).

9
Effect of Stock Holdings Based on Buy-Side Analysts Recommendations5

Prior research documents what we term a stock holding effect in that investors with a

long position (i.e., a holding) in a stock tend to assign higher earnings estimates than those with a

short position in an identical firm (Hales 2007). Han and Tan (2010) show that investors with a

stock holding (long position) tend to make higher earnings estimates compared to those with no

stock holding (a neutral position) when given positive management guidance provided in the

form of range guidance. This stock holding effect has also been documented among sell-side

analystsstock recommendations originating from sell-side analysts are more optimistic if the

stock is held by the brokerage firms mutual fund clients and particularly so when trading

commissions paid by the clients increase (Firth, Lin, Liu, and Xuan 2013).

Prior research (e.g. Hales 2007; Han and Tan 2010) maintains that the psychological

mechanism responsible for this stock holding phenomenon is related to a directional preference

effect, where the desire for a preferred outcome influences ones assessment of the likelihood of

that event occurring (Krizan and Windschitl 2007). This effect is also broadly related to the

phenomenon of motivated reasoning (Kunda 1990), where individuals unconsciously interpret

and process information in a manner that fits in with their directional goals, subject to

reasonableness constraints (for examples of accounting studies, see Eames, Glover, and Kenney

2002; Kadous, Kennedy, and Peecher 2003; Wilks 2002). In the context of our study, buy-side

analysts are employees of the asset management firm and it is conceivable that the firms stock

holding in itself is sufficient to induce unconscious motivated reasoning among these analysts

when the firm has a stock holding.

5
We develop our theory below with reference to buy-side analysts. Similar arguments apply for fund managers. For instance, the
fact that an asset management firm has a holding in a stock per se does not lead fund managers to have directional preferences for
the stocks earnings and price performance to improve. The fund manager has to be the one who made that recommendation.
Similarly, fund managers (particularly small cap fund managers; see Frey and Herbst 2014) do periodically make stock
recommendations and earnings forecasts that are accessible to other fund managers and selected clients.

10
A counter-point is that the buy-side analyst setting is distinct from an investor or sell-side

analyst setting in that the buy-side analyst does not have a direct holding in a stock (as with

investors) or benefit from increased commissions when clients with larger stock holdings trade

more (as with sell-side analysts). 6 Hence, the funds holding of a stock may not, by itself, evoke

motivated reasoning by buy-side analysts. On the other hand, a distinctive institutional feature in

the buy-side analyst environment is that an asset management firms asset and fund allocation

(and, therefore, stock holdings) are significantly influenced by in-house buy-side analysts

recommendations (Cheng et al. 2006; Frey and Herbst 2014), which means that if the asset

management firm has a stock holding, it likely originated from a recommendation by the in-

house buy-side analyst.7

Fund managers we spoke to mentioned that within an asset management firm, there is

stiff internal competition among buy-side analysts for stock ideas to be adopted by fund

managers, 8 and therefore, pressure for the adopted stock to perform well. This competition for

fund managers attention and asset allocation implies that a buy-side analysts recommendation

may not translate into a position by an asset management firm. It also suggests that once a stock

recommendation is adopted by the asset management firm and translated into a stock holding,

the reputational stakes are high for the buy-side analyst.

6
In the case of sell-side analysts, their brokerage firms earn commissions from their clients and sell-side analysts rewards are
tied to this commission income. Hence, they have direct financial incentives to optimistically bias their reports to please their
clients who already hold the stock. Optimistic reports sustained over time also encourage more trading by the clients, and clients
with large stock holdings who trade more directly increase the analysts commission income. Unlike sell-side analysts, buy-side
analysts do not directly benefit from trading by the fund managers (Cheng et al. 2006).
7
Our fund manager contacts also confirmed this, and further indicated that within an asset management firm, a single buy-side
analyst generally covers a particular stock (see also Groysberg et al. 2008). We surveyed 49 fund managers and asked whether, in
their asset management firms, fund managers stock holdings result from stock recommendations by internal buy-side analysts.
About 86 percent of them indicate that this is the case.
8
This competition likely explains why a majority of buy-side analysts forecasts involve buy/hold recommendations. Frey and
Herbst (2014) find that for the dataset of forecasts issued by buy-side analysts from a global asset management firm, a majority of
recommendation revisions involve an upgrade to hold or buy recommendations. Frey and Herbst (2014) document that over 73
percent of all recommendation revisions start from at least a hold recommendation, and their Table 2 indicates that approximately
83 percent of all recommendation revisions involve an upgrade.

11
Thus, buy-side analysts can have a vested interest in the earnings and performance of the

company under coverage if the asset management firm has made an investment in the stock

through their recommendations. Specifically, two conditions are necessary: (a) the asset

management firm has made an investment in the stock, and (b) the investment is based on the

buy-side analysts recommendation. Both conditions are inter-related in the asset management

environment, and these features imply that two competing mechanisms may be at play here.

The first mechanism is motivated reasoning (Kunda 1990). Given the link between the

asset management firms stock holding and the buy-side analysts recommendation, the career

consequences of bad recommendations strengthen the personal stakes for the buy-side analysts

and therefore their preferences for the stock to do well. Motivated reasoning predicts that this

stock-holding-plus-recommendation effect can bias their earnings forecasts upwards. An

important aspect of this motivated reasoning mechanism is that it is unconscious (Kunda 1990).

The second mechanism relates to a strategic justification effectan impression

management tactic aimed at altering the recipients impression of the event (Schlenker and

Weigold 1992). Research on impression management indicates that decision makers actively use

different means to seek to influence significant others impression of themselves (see Leary and

Kowalski 1990 for a review), and this results in strategic endorsement or justification of specific

attitudes (Gaes, Kalle, and Tedeschi 1978) and escalating commitment to a failing course of

action (Brockner, Rubin, and Lang 1981; Staw 1981; Schoorman and Holahan 1996). The

motivation to manage the impressions of others is strong when the decision maker is personally

identified with and held responsible for the decision and has his/her reputation staked on its

success. For instance, Caldwell and OReilly (1980) documented how responsibility for a failed

outcome led to greater impression managementin their study, decision makers held responsible

12
for a failed outcome engaged in impression management by strategically presenting more

favorable information to their superiors. In our current setting, the buy-side analyst is responsible

for the firms stock holding (having made the stock recommendation), and his/her reputation is

staked on the stock performing well. Issuing more optimistic earnings forecasts signals to

important stakeholders that the buy-side analyst is confident about the stocks prospects. It is

also a way to justify his/her buy recommendation.9 Consistent with this argument, an archival

study by Eames et al. (2002) finds evidence that analysts bias their forecasts upwards/downwards

in a direction that supports their buy/sell recommendations. 10 Bradshaw (2002) also finds

evidence consistent with the argument that sell-side analysts use target prices to justify their

stock recommendations. There are two important aspects and assumptions in the strategic

justification mechanism: (a) the justifications (whether in the form of earnings forecasts,

recommendations or target prices) have to be publicly seen by the target audience whose

impressions the analysts desire to manage, and (b) the resulting bias is conscious.

Effect of Stock Holding via Analysts Recommendation: Public versus Private Disclosure

The different assumptions made by the two competing mechanisms permit us to make

predictions and test which mechanism holds in the buy-side analyst institutional setting. The

strategic justification mechanism involves the need to impression manage, and requires that the

buy-side analysts forecasts be disseminated publicly; it is a conscious process. In contrast, the

motivated reasoning mechanism is an unconscious process that involves a directional preference

for the stock to do well (for instance, earnings to be higher than consensus). Since the

9
This strategic justification effect does not apply to the earlier setting when we consider the mere effect of the asset management
firm having a stock holdingin that setting, we do not consider the case where the holding is based on the analysts
recommendation, and so, the issue of the analyst having a reputational stake in the successful outcome of the recommendation
does not arise.
10
Eames et al. (2002) argue that their results are consistent with an unconscious motivated reasoning rather than an intentional
trade boosting argument. However, their results are also consistent with an intentional strategic justification effect that we discuss
here.

13
unfavorable consequences result whether the forecasts are publicly disseminated or not,

motivated reasoning should occur in both public and private dissemination settings. We

elaborate on this below and discuss our predictions in the public setting followed by the private

dissemination setting.

The public dissemination setting is a naturally occurring feature in the buy-side analyst

environment in that the analysts forecasts are made public to other analysts and fund managers

in the firm, along with clients. Buy-side analysts forecasts and recommendations are issued with

the express purpose of selling stock ideas to the asset management firm and its clients

(Groysberg et al. 2008).11 Portfolio managers we interviewed indicated that, for marketing and

client relationship purposes, these buy-side analysts recommendations/forecasts are

disseminated to fund managers and selected clients of the asset management firm.12

In this public dissemination setting, suppose that, after the asset management firm had

made a significant fund allocation to a stock based on a buy-side analysts recommendation,

management issues guidance that is lower than consensus forecasts. The motivated reasoning

mechanism maintains that because buy-side analysts prefer that their recommended stock to have

good future prospects (as this has career consequences for them), they will issue more optimistic

forecasts, albeit unconsciously. The strategic justification mechanism makes a different argument

but predicts the same effect. Given the public dissemination, fund managers and clients of the

asset management firm, who are important stakeholders to the buy-side analyst, are likely to be

unhappy about the professional advice they received from the buy-side analyst should they

subsequently receive pessimistic earnings forecasts about the stock. The strategic justification

11
These recommendations (and the associated earnings forecasts) are likely buy recommendations because asset management
firms generally have restrictions on short-selling (Groysberg et al. 2008). For instance, Frey and Herbst (2014, Table 2) indicate
that only 14 percent of buy-side analysts recommendation revisions involve a downgrade to underperform or sell, and 73
percent of the revisions start from a hold or buy position.
12
The webpage by Graeme Pietersz, former equity analyst, makes a similar point about buy-side analysts reports being made
available to clients (http://moneyterms.co.uk/buy-side-analyst).

14
mechanism maintains that in this public dissemination setting, the buy-side analyst has a high

need to impression manage and justify the appropriateness of his/her prior recommendation,

which intensifies his/her defensive bolstering as to why the recommendation is correct (Lerner

and Tetlock 1999). Consciously issuing more optimistic forecasts is a way to achieve this.

Thus, both mechanisms predict a stock-holding-plus-recommendation effect in this public

dissemination setting, which makes it impossible to distinguish the two mechanisms. However,

these mechanisms make different predictions in a somewhat hypothetical private dissemination

setting that we can create in an experimental setting, where the buy-side analysts make private

forecasts known only to themselves.

Motivated reasoning is unconscious (Kunda 1990), so one would expect an effect of the

asset management firm holding the stock based on the buy-side analysts recommendation even

in a private forecast setting. The reputational and career consequences of a bad recommendation

equally apply, whether the forecasts are made publicly or privately. If this concern about the dire

reputational and career consequences looms large such that it has unconscious effects on the buy-

side analysts judgments, the stock-holding-plus-recommendation effect should apply in both

private and public forecast dissemination settings. 13 In contrast, the strategic justification effect

is a conscious effect arising from the need to justify ones prior actions to others, and in our

setting, issuing more optimistic forecasts for the consumption of stakeholders is one way to do so.

In contrast, when a forecast is kept private, there is no need to upwardly-bias earnings forecasts

13
It is conceivable that holding effects are stronger in a public (versus private) dissemination setting because of
stronger directional preferences, or remain equally strong across public and private settings because the directional
preferences are already present in a private dissemination setting. All the studies on motivated reasoning that we
are aware of use settings where there is no public dissemination of the participants judgments, akin to a private
dissemination setting we use here (see Hales 2007, Han and Tan 2010 for accounting examples). The
theoretical mechanism argues that this is an unconscious process that should apply even when there is no
concern that the forecast provided will be seen and evaluated by a significant other.

15
when there is a stock holdingthese forecasts will not be seen by others to begin with; in other

words, there is no stock holding effect. In essence, these private forecasts are the buy-side

analysts true beliefs about the stocks prospects.

Predicting a stock holding effect in a setting where the analyst privately expresses his/her

true beliefs does require a relatively strong assumption about the pervasiveness of this stock

holding bias. Survival in the highly-competitive buy-side analyst labor market requires these

analysts to be able to separately generate private beliefs about a stocks prospects that are

stripped of incentive-driven bias, so it is possible that buy-side analysts are able to purge

themselves of an unconscious stock-holding-plus-recommendation effect, as driven by motivated

reasoning, such that we do not predict this effect in a private forecast disclosure setting. Further,

while prior demonstrations of directional preferences effects involve personal holdings of a stock

(e.g., Hales 2007; Han and Tan 2010), the buy-side analyst setting involves one where the stock

holding is indirect (via the company the analyst is working for), notwithstanding an involvement

by the analyst in that the firms holding arises from the analysts stock recommendation. Thus,

the motivated reasoning effect may be weaker in this setting. Overall, we anticipate that this

weaker form of the stock holding bias will be more likely, consistent with the strategic

justification mechanism. Referring to the graphical depiction (Figure 1) of our ordinal interaction

prediction, this suggests a positive stock-holding-effect slope in the public forecast condition,

and a flat line (no stock holding effect) in the private forecast condition. There is no incentive or

preference by buy-side analysts for the stock to perform in the no-holding condition, so we do

not anticipate any effect of public versus private forecasts here. This also suggests that the

combination of public dissemination and stock holding induces the most optimistic forecasts

16
among buy-side analysts, and no difference across the other conditions. Our hypothesis is stated

below:

Hypothesis 1. When buy-side analysts earnings forecasts are made public, their forecasts will
be higher when the asset management firm holds a stock made through their recommendation
than when it does not do so, but their forecasts will not be induced higher by the firms stock
holding when their forecasts are kept private.

We note that our theory and design allow us to determine which mechanism is in place.

In a public forecast dissemination setting, we expect a positive slope in Figure 1, based on both

motivated reasoning and strategic justification mechanisms. In a private forecast dissemination

setting, a positive but less steep slope is consistent with the motivated reasoning mechanism,

while a flat line is consistent with the strategic justification mechanism.

[Insert Figure 1 here]

RESEARCH METHOD

Participants

Our participants were fund managers and buy-side analysts from 13 international firms

and 14 major Taiwanese firms.14 We include both fund managers and buy-side analysts in our

sample because fund managers are a subset of buy-side analysts (see footnote 5), and most of the

fund managers in our sample have prior experience as buy-side analysts. For brevity, we

collectively term our participants buy-side analysts. One of the authors contacted these firms

senior executives and obtained their agreement to provide participants. We then sent the research

instruments to these executives, and asked them to distribute the instruments to their staff for

completion. In total, we distributed 440 copies of the instruments and received 250 usable

14
International firms are branches or subsidiaries of asset management firms whose major shareholders are not Taiwanese firms
or individuals; local firms are those whose major shareholders are Taiwanese firms or individuals. Both types of firms follow
similar practices.

17
responses (response rate of about 57 percent). We omitted 19 respondents as their earnings

estimates were outliers, being more than three standard deviations apart from the other responses

and/or erroneous (annual EPS less than sum of quarterly EPS). Among the final sample of 231

usable responses from participants, 154 were fund managers (average of 5.6 years as fund

manager, 5.9 years previously as buy-side analyst, 10.4 years in finance), 63 were buy-side

analysts (average 3.5 years as buy-side analyst), 5 indicated others, and 9 did not include their

designations.15,16 On average, the scale of funds managed by these firms and participants added

to US$38 billion and US$179 million, respectively. The average number of employees in the

firms was 108 and the participants average length of experience in finance was 8.6 years.

Design

Our design was a 2 x 2 + 2 between-subjects design. The first factor we manipulated was

Holdingin the Holding condition, consistent with the institutional features discussed earlier,

participants were told that the asset management firm had invested a significant portion of its

funds in a target stock based on the participants prior recommendation. In the No Holding

condition, they were told that the firm had not made any fund allocation to the stock.17 The

second factor we manipulated was Dissemination (private, public), which varied whether the

buy-side analysts forecasts would be kept private for their own use or released to others

15
Our results are very similar and tests of the holding-plus-recommendation effect remain significant when we separately analyze
the data using only the fund manager sample. Our results are directionally similar but become insignificant if we only use the
buy-side analyst sample as this is a much smaller sample (with n = 6 in some conditions).
16
Among these participants, 59% are from Taiwanese firms and 41% from international firms. We also included firm type (local,
international) as a third independent variable. Our results for our hypothesis tests are very similar, and firm type does not
moderate our results (F(1, 128)=0.023, p = 0.880). Our results are also not influenced by other demographic background such as
experience or firm size.
17
In the No Holding condition, we do not indicate that the asset management firm has no stock holding based on the buy-side
analysts recommendation to sell (or not to hold) the stock. Once the firm has no holding in a stock (and there is subsequently bad
news related to this stock), the analysts directional preferences for the stock to perform well or strategic justification effects will
be lower whether the analyst had made a prior recommendation not to hold the stock. Importantly, when an asset management
firm does not currently have a stock holding, it is less likely that this is because the buy-side analyst has made a sell
recommendation. For instance, Frey and Herbsts (2014) find that the majority of all recommendation revisions involve upgrades
(see footnote 11).

18
(including colleagues and clients). The Private Dissemination condition, where the buy-side

analysts forecasts are known only to themselves, is one that does not exist in the institutional

setting of asset management firms (although it is plausible that buy-side analysts form private

forecasts that they then adjust before making these forecasts public). We capitalize on the

comparative advantage of the experimental method to include this condition which serves as a

clean baseline to assess the effect of public dissemination of the analysts forecasts, one that is

not possible using archival data. We also included an Accuracy Goal condition where

participants were instructed that their goal was to be as accurate as possible in their earnings

forecasts; no information on stock holding or dissemination of their forecasts was provided.

Finally, to disentangle the incremental effect of having made a personal recommendation to hold

the stock from the firm having a stock holding, we also incorporated a modified Holding/Public

condition which was identical to the Holding/Public condition except that participants were not

told that the stock holding arose out of their stock recommendation.

Procedure

Our instrument was adapted from Han and Tan (2010), and modified based on a pilot test

and interviews with senior fund managers. In the experimental materials, we provided all

participants with identical background information about the industry, products, markets and

competition of the target company (Kappa, Inc.). We also gave all participants the companys

history of earnings in the past three years, including quarterly and annual earnings in year X1

and year X2, and quarterly earnings for the first and second quarters of year X3. They were also

provided with the consensus forecasts for the third quarter ($0.25) and the whole ($0.80) of year

X3 that were announced after the second quarter of year X3. In addition, participants received

managements earnings guidance for the third quarter of year X3 ($0.16 to $0.20), which was

19
below consensus forecast. We use this bad news setting so that effects of directional

preferences or strategic justification would be apparent in that these mechanisms would result in

more optimistic forecasts, whereas the news would suggest otherwise. The effects of these

mechanisms would be directionally similar in a good news setting where guidance is above

consensus forecast.

Participants in the experiment were asked to make a forecast of the target companys

earnings. Our instructions to participants after the provision of managements guidance for X3

quarter 3 varied depending on the experimental condition they were assigned to.18 In the Holding

condition, participants were told to assume that just before Kappas announcement, your firm

has allocated a fairly significant portion of its funds in Kappa based on your recommendation.

In the No Holding condition, they were told to assume that their firm has not made any fund

allocation to Kappa. In the Private condition, they were asked to assume that the forecasts that

they made will be kept private for their own use only, and not released to others, including

colleagues and clients. In the Public condition, they were asked to assume that the forecast that

they made will be released to others, including colleagues and clients. After reading these

instructions, all participants were asked to provide Kappas EPS forecasts for the third quarter of

year X3, full-year X3, and full-year X4. After they had made the forecasts, all participants

answered a post-experimental questionnaire, including demographic and debriefing questions.

We discuss details of the Accuracy and modified Holding conditions later.

RESULTS

18
The guidance and manipulation instructions appear on the same page, with the guidance appearing before the manipulations
instructions. This mimics a real-world situation where buy-side analysts read/hear about management guidance in the
press/conference calls, and are immediately reminded of their prior stock recommendations.

20
As a check on our Holding manipulation, we asked participants to indicate the amount of

funds their firm had allocated to Kappas stock, with the options being either no earlier fund

allocation or a significant fund allocation to Kappa based on their previous recommendation. The

modified Holding condition contained the same manipulation check but without the words

based on your previous recommendation. About 78 percent of the respondents answered this

question correctly. As a check on our Dissemination manipulation, we asked participants to

indicate whether they assumed their earnings forecasts would be released to others (including

colleagues and clients) or kept for their own use. About 68 percent of the participants correctly

answered this question. Subsequent debriefing with some of the participants indicated that they

understood the manipulations, but misinterpreted the debriefing questions. We also engaged in

an extensive debriefing exercise with a sample of 30 participants at the end of our session, after

participants had returned the instruments. Specifically, we interviewed these participants to

verify that they understood and correctly interpreted the manipulations. All of them did. As such,

our analyses that follow include all participants.19

In our experimental materials, the most recent update on future earnings came from the

third quarter earnings guidance from management, who provided below-consensus guidance for

that quarter. Our primary dependent measure is, therefore, participants third quarter earnings

estimates.20 Given our ordinal interaction prediction associated with our hypothesis, we test this

19
Our results become either marginally significant or insignificant when we drop participants with manipulation check failures.
20
As with the Q3 forecasts, for the full-year X3 forecasts, we also find no significant effects of holding (versus no holding),
given private dissemination (means=80.5 and 83.3, respectively; t132=0.855, p=0.394), and a significant effect of holding (versus
no holding), given public dissemination (means=82.6 and 76.6, respectively, t132 = 2.030, p = 0.022, one-tailed). The two-way
interaction using the conventional ANOVA test is significant (F(1,132) = 3.984, p = 0.048), but not the overall contrast test
(F(1,132) = 1.051, p = 0.307), suggesting that the full predicted pattern is not supported. For full-year x4 forecasts, none of these
comparisons or tests is significant (smallest p = 0.329). The stronger results for the quarter 3 forecasts compared to annual
forecasts likely occurred because in our experiment, management provided guidance (in the form of a range) only for quarter 3
but not full-year earnings, and thus, the quarter 3 guidance served as the strongest basis for the analysts to make their forecasts.
The study by Han and Tan (2010), which uses a similar setting, also finds significant results with quarter 3 forecasts but not full-
year forecasts. An alternative explanation may be that the buy-side analysts deliberately attempted to bias only short-term but not
long-term earnings.

21
hypothesis using contrast-coded ANOVA to maximize the power of our tests (Buckless and

Ravenscroft 1990; Rosnow and Rosenthal 1995; see also Sedor 2002 and Seybert 2010 for

applications in accounting settings). Table 1, panel A shows the descriptive statistics. Table 1,

panel B reports the ANOVA results, which shows a marginally significant Holding x

Dissemination interaction (p=0.062, one-tailed). Table 1, panel C shows the contrast-coded

ANOVA results, while Table 1, panel C reports tests of simple main effects. Figure 2 illustrates

the actual results.

Hypothesis 1 predicts that when buy-side analysts earnings forecasts are made public,

their forecasts will be higher when the asset management firm holds a stock made through their

recommendation than when it does not do so, but their forecasts will not be induced higher by

the firms stock holding when their forecasts are kept private. We perform an overall test for this

ordinal interaction using contrast weights of 3 for the Public/Holding condition, and -1 for the

remaining conditions. The contrast-coded ANOVA results show that the contrast is statistically

significant (F(1, 132) = 3.476, p = 0.032, one-tailed; Panel B of Table 1). The residual mean

square, denoting the between-subjects variance unexplained by the contrast effect, is

insignificant (F(2, 132) = 0.043, p = 0.957), which implies that there is no significant between-

condition variance unexplained by our a priori contrast.

As further analyses, we conduct tests of simple main effects (see Panel C of Table 1).

Results show that when the forecasts are made public, participants quarter three forecasts are

significantly higher when their firm holds the stock (mean = 20.38 cents) than when it does not

(mean = 19.17 cents; t132 = 1.699, p = 0.046, one-tailed). Participants forecasts are not

significantly different whether the fund holds the stock (mean = 19.08 cents) or not (mean =

19.56 cents; t132 = 0.606, p = 0.545) when the forecasts are kept private. This result is consistent

22
with the strategic justification effect, which predicts no effect when forecasts are kept private, 21

and inconsistent with the motivated reasoning effect, which predicts an effect here.

[Insert Table 1 about here]

[Insert Figure 2 about here]

We also analyze the simple main effects of public versus private dissemination. Table 1,

panel C shows that consistent with our theory, in the Holding condition, participants forecasts

are higher when the forecasts are made public (mean = 20.38 cents) than when they are kept

private (mean = 19.08 cents; t132 = 1.684, p = 0.047, one-tailed). However, when the asset

management firm does not hold the stock, forecasts that are made public or kept private are not

significantly different (means = 19.17 cents and 19.56 cents; t132 = 0.528, p = 0.599). This latter

result is broadly consistent with the idea that the positive effect on forecasts associated with

making the forecasts public will be attenuated in the no-holding condition.22

Accuracy Goal Condition: Evidence of Bias in the Public/Holding Condition

A premise we make is that the Public/Holding condition is biased, or at least, more biased

than the other conditions. To verify that a bias exists, we create an Accuracy Goal condition

where participants were told that their goal was to be as accurate as possible in their earnings

forecasts. They were not informed of any stock holding position by their firm, or whether their

forecasts would be communicated to others (or kept private). The mean Q3 earnings forecasts of

21
We coded participants rationale for their earnings forecasts, and computed a Netpositive measure by taking the number of
positive issues minus the number of negative issues mentioned. We find no significant main or interaction effect (p > 0.175).
However, an a priori contrast test indicates that Netpositive is higher in the Public/Holding condition than the average of the
other three conditions (p = 0.038, one-tailed). Netpositive is also higher in that condition than each of the other conditions (p <
0.100, one-tailed) except for the Public/No Holding condition (p = 0.108, one-tailed).
22
We also measured the extent to which participants recommend the stock and the extent to which they believe the stock price
will appreciate in the future. Our manipulated variables do not interactively influence these measures. However, we find evidence
that their quarter three earnings estimates influenced their stock recommendation and stock price appreciation potential
judgments. Both these measures are correlated with their quarter three earnings estimates (r > 0.31, p = 0.000). Furthermore, with
either of these measures as the dependent variable, ANCOVA with our manipulated variables as independent variables and
quarter three earnings estimates as covariate indicates that the coefficient on quarter three earnings is statistically significant (p =
0.000). The interaction term involving our manipulated variables becomes statistically insignificant (p > 0.426).

23
the 25 participants is 18.72 cents, which is lower than that in the Public/Holding condition (t156 =

2.118, p = 0.018, one-tailed), but not significantly different from any of the other conditions (p >

0.297).23 This analysis provides evidence that the Public/Holding condition leads to upwardly-

biased or optimistic earnings forecasts.

Disentangling the Incremental Effect of Having Made a Personal Recommendation from

Having a Stock Holding

In this section, we investigate whether stock holding per se induces forecast optimism.

Because of data constraints, we restrict this investigation to the Holding/Public condition, one

where we expect the largest forecast optimism. We do not examine the incremental effect of

holding per se in a Private setting because we expect the effect of holding (in the presence of a

prior recommendation) to be small to begin with.

In the Holding condition, we instructed participants to assume that the firm had invested

a significant amount of funds in Kappa based on their recommendation. To directly assess

whether informing participants that the firm has a holding in a stock (without specifying that this

comes from reliance on their stock recommendation) influences participants directional

preferences, we included a modified Holding/Public condition. In addition to being informed that

their forecasts will be released to others (including colleagues and clients), participants in this

condition are also informed about the following: Assume that just before Kappas

announcement, your firm has allocated a fairly significant portion of its funds in Kappa. We

restricted this modified Holding condition to the Public condition because the Public (relative to

the Private) condition is the one where we expect the larger effect of stock holding.

23
Specifically, the mean Q3 earnings forecast in the Accuracy Goal condition is not significantly different from that in the
Private/No Hold condition (t156 = 1.046, p = 0.297), the Private/Hold condition (t156 = 0.415, p = 0.678), or the Public/No Hold
condition (t156 = 0.562, p = 0.575).

24
Participants (n = 70) 24 mean Q3 forecast in the modified Holding/Public condition is

18.73 cents, significantly lower than the 20.38 cents forecast in the Holding/Public condition (t225

= 2.246, p = 0.026). This forecast in the modified Holding/Public condition is not different from

the 19.17 cents forecast in the No Holding/Public condition (t225 = 0.580, p = 0.563) or the

Accuracy Goal condition (t225 = 0.010, p = 0.992). Thus, in the absence of information about a

direct involvement in making the stock recommendation, stock holding per se, even with public

dissemination of the forecasts, does not lead to bias.25

These results are informative from a theoretical perspective in that it shows that it is not

the firms stock holding per se that creates the directional preferences for the buy-side analyst,

but the incremental effect of personal recommendation leading to the firms stock holding that

does so. 26

Supplementary Experiment

To provide more direct evidence that participants consciously provide different forecasts

under the two dissemination conditions, we conducted another experiment with twenty-one fund

managers and 4 buy-side analysts. Their average experience in finance is 9.68 years. The average

experience as fund managers and buy-side analysts is 5.09 years and 5.98 years, respectively.

The fund size managed by their firms and by the participants is US$26 billion and US$82

24
During the administration of the experiment, we inadvertently distributed more copies of this instrument than the others, and
correspondingly received more responses back.
25
The fact that we find no effect of stock holding alone in the Public condition (one where we expect stronger effects compared
to the Private condition) also suggests that stock holding alone is unlikely to have an effect in the Private condition.
26
In the post-experiment section, we assessed participants general preferences for private versus public release of their forecasts,
and asked them to respond to questions about another hypothetical company based on their general experiences. Specifically, for
one question, we asked participants to assume that they had estimated Company As earnings per share to be above consensus
forecast based on managements guidance, and to indicate whether they preferred that their earnings estimates for Company A
were kept private or released to others (including colleagues and clients). They answered the question on a 15-point scale (-7:
definitely prefer my estimates to be kept private, +7: definitely prefer my estimates to be released to others). We then asked them
a second question which was identical to the first question except that they were to assume that they had estimated Company As
earnings to be below consensus. The mean responses to the two questions are 0.196 and -0.226, respectively, with the difference
being statistically significant (pairwise difference t132 = 2.509, p = 0.013). Thus, participants have a preference for their forecasts
to be kept private if they issued below-consensus earnings rather than above-consensus forecasts. This result is consistent with
participants being aware of the self-presentation effects of public versus private dissemination of their forecasts.

25
million, respectively. While the size of funds managed by the participants and their firms is

smaller than that in the main experiment, the experiences of the participants are about the same

across the two experiments.

In the experiment, we used a within-subjects design where participants either provide

forecasts for the Private condition followed by the Public condition (n=12), or vice versa (n=13).

We provided the participants with the same background information about Kappa Inc. as in the

main experiment. We only asked them to provide the earnings forecast for Kappa, Inc. for the

third quarter of Year X3, unlike the main experiment where we also asked them to provide full

year forecasts and debriefing questions.

Both within-subjects conditions involve the firm having a stock holding based on the

prior recommendation of the analyst, the setting where both motivated reasoning and strategic

justification mechanisms predict an effect in a between-subjects setting. In this within-subjects

setting, however, the motivated reasoning mechanism predicts no difference in this within-

subjects settingsince the effect is unconscious, it should disappear once the public/private

difference is made obvious in the within-subjects design. In contrast, the strategic justification

mechanism is a conscious process, and predicts higher forecasts in the Public condition than in

the Private condition even in a within-subjects setting.

The results indicate a significant main effect of Dissemination, with the mean forecast

under in the Public condition higher than that in the Private condition (means = 0.1932 and

0.1904, respectively; F (1, 23) = 3.520; p = 0.0365, one-tailed). The interaction with order is not

significant (F (1, 23) = 0.116; p = 0.368, one-tailed). This result is consistent with participants

intentionally and strategically offering higher forecasts in the Public (versus Private) setting,

consistent with the strategic justification mechanism. The motivated reasoning mechanism would

26
have predicted no difference in this within-subjects setting (since the effect is unconscious, and

the within-subjects design make the public/private difference obvious to the participants).

CONCLUSION

In this study, we conduct an experiment with buy-side analysts and fund managers where

we manipulate whether the mutual fund has made an investment in the stock based on the

recommendation of the analyst, and whether the forecasts are released to colleagues and clients

or kept private for their own use. While an asset management firms stock holdings are

significantly influenced by buy-side analysts recommendations and earnings forecasts, our

results suggest that the reverse is, paradoxically, also truebuy-side analysts forecasts can be

influenced by the very stock holdings that the firm holds, particularly when these forecasts are

disseminated to other fund managers and external clients. We also note that while buy-side

analysts forecasts are not publicly disseminated to the market like those made by sell-side

analysts, the parties receiving buy-side analysts forecasts are targeted and important

stakeholders (e.g., fund managers who evaluate them). Further, while not all clients receive the

asset management firms forecasts for various stocks, selected clients do and they are generally

important current/potential clients (private communication with fund managers).

Our study provides the first demonstration of a causal link between stock holdings based

on a buy-side analysts recommendation and forecast optimism, as moderated by public/private

dissemination of buy-side analysts forecasts. Our investigation provides a better theoretical

understanding of how two institutional features within an asset management firm context lead to

optimism in buy-side analysts forecasts. Our finding that it is the Holding/Public dissemination

condition that leads to optimistic forecasts has implications for asset management firms in terms

of measures they can adopt to reduce forecast optimism among their analysts. There is probably

27
little that asset management firms can do to prevent dissemination of buy-side analysts forecasts

within the firm, given that these forecasts are intended for use within the firm. There is also

probably little the firm can do to avoid the effect of stock holding because buy-side analysts

necessarily cover the stocks that the firm holds. Further, our interviews with senior fund

managers and the results in Frey and Herbst (2014) indicate that asset management firms stock

holdings generally result from stock recommendations from buy-side analysts, which imply that

in practice, these stock holdings will induce forecast optimism among buy-side analysts (because

they are generally involved in the firm taking a holding in the stock). However, our results on the

modified public dissemination/holding condition, where participants are merely told of a stock

holding but not an involvement through having made a stock recommendation, is instructive.

The results indicate that it is this involvement in having made a recommendation that is an

important factor in inducing optimism in their forecasts. Further, our results showing that it is an

intentional bias suggest that removing the incentives to be biased will be beneficial. An

implication, therefore, is that asset management firms may benefit from having more than one

buy-side analyst cover the same stock. Our understanding from fund managers is that a common

industry practice is for a single buy-side analyst to cover a single stock, which likely exacerbates

the bias we document in this paper. One possibility then is for the asset management firm to

engage another analyst to cover the stock (in addition to the one who had made the

recommendation) once a stock holding has been made. The second analyst, not having made the

initial recommendation, is likely to be more objective. Another possibility is to have periodic

rotation of analysts covering a particular stock. These measures do entail costs to the firm and/or

analyst.

28
Our study is subject to some limitations. One limitation is that because of participant

constraints, we do not have a full factorial design where we systematically cross whether the

analyst has made a stock recommendation with the firms stock holding across the forecast

dissemination conditions. Thus, while we know the incremental effect of having made a stock

recommendation that leads to a stock holding when there is public dissemination of the forecasts,

we do not know its effect with private dissemination. Our expectation though is that given the

absence of an effect of stock holding with recommendation when there is private dissemination,

the sole effect of recommendation itself is unlikely to be significant. In addition, we do not know

whether the act of having made a recommendation (absent holding) is sufficient to induce

optimistic forecasts with public dissemination. While this issue is of theoretical interest, our

discussion with fund managers indicate that it is relatively rare for asset management firms to

publicly release forecasts of stocks that they do not already hold. Also, we examine a context

involving a single stock. In practice, buy-side analysts cover several stocks and it is possible that

an ill-timed recommendation for a single stock does not matter if the other stocks recommended

in their portfolio are doing well. Suffice to say, buy-side analysts work in an institutionally rich

environment. Our study identifies two integral institutional aspects of asset management firms

that influence buy-side analysts forecasts, and it is fruitful for future research to examine the

impact of other environmental factors on these forecasts.

29
REFERENCES

Affleck-Graves, J., L. Davis, and R. Mendenhall. 1990. Forecasts of earnings per share: Possible
sources of analyst superiority and bias. Contemporary Accounting Research 6: 501517.
Ashton, R. H., and A. M. Cianci. 2007. Motivational and cognitive determinants of buy-side and
sell-side analyst earnings forecasts: An experimental study. The Journal of Behavioral
Finance 8 (1): 919.

Boston Consulting Group. 2010a. In search of stable growth. Boston, MA: Boston Consulting
Group.

Boston Consulting Group. 2010b. Building on success: Global asset management 2011. Boston,
MA: Boston Consulting Group.

Bradshaw, M. T. 2002. The use of target prices to justify sell-side analysts stock
recommendations. Accounting Horizons 16 (1): 2741.

Bradshaw, M. T. 2011. Analysts forecasts: What do we know after decades of work? Working
paper. Boston College.

Brockner, J., R. Z. Rubin, and E. Lang. 1981. Face-saving and entrapment. Journal of
Experimental Social Psychology 17(1): 68-79.

Buckless, F. A., and S. P. Ravenscroft. 1990. Contrast coding: A refinement of ANOVA in


behavioural analysis. The Accounting Review 65 (4): 933945..
Caldwell, D. F.,and C, OReilly. 1980. Response to failure: The effects of choice and
responsibility on impression management. Academy of Management Journal 25: 121-
136.
Cheng, Y., M. H. Liu, and J. Qian. 2006. Buy-side analysts, sell-side analysts, and investment
decisions of money managers. Journal of Financial and Quantitative Analysis 41 (1): 51
83.

Cowen, A., B. Groysberg, and P. Healy. 2006. Which types of analyst firms make more
optimistic forecasts? Journal of Accounting and Economics 41: 119146.

30
Eames, M., S. M. Glover, and J. Kennedy. 2002. The association between trading
recommendations and broker-analysts earnings forecasts. Journal of Accounting Research
40 (1): 85104.

Firth, M., C. Lin, P. Liu, and Y. Xuan. 2013. The client is king: Do mutual fund relationships bias
analyst recommendations? Journal of Accounting Research 51 (1): 165200.
Frey, S., and P. Herbst. 2014. The influence of buy-side analysts on mutual fund trading. Journal
of Banking and Finance 49: 442458.
Gaes. G. G., R. J., Kalle, and J. T. Tedeschi. 1978. Impression management in the forced
compliance situation: Two studies using the bogus pipeline. Journal of Experimental
Social Psychology 14: 493-510.
Groysberg, B., P. Healy, and C. Chapman. 2008. Buy-side vs. sell-side analysts' earnings
forecasts. Financial Analysts Journal 64 (4): 2539.
Groysberg, B., P. Healy, and G. Serafeim. 2013. The stock selection and performance of buy-side
analysts. Management Science 59 (5): 10621075.
Hales, J. 2007. Directional preferences, information processing, and investors' forecasts of
earnings. Journal of Accounting Research 45 (3): 607628.
Han, J., and H. T. Tan. 2010. Investors' reactions to management earnings guidance: The joint
effect of investment position, news valence, and guidance form. Journal of Accounting
Research 48 (1): 81104.
Hopkins, P. E., R. W. Houston, and M. F. Peters. 2000. Purchase, pooling, and equity analysts'
valuation judgments. The Accounting Review 75 (3): 257281.
Kadous, K., S. J. Kennedy, and M. E. Peecher. 2003. The effect of quality assessment and
directional goal commitment on auditors' acceptance of client-preferred accounting
methods. The Accounting Review 78 (3): 759778.
Ke, B., and Y. Yu. 2006. The Effect of Issuing Biased Earnings Forecasts on Analysts Access to
Management and Survival. Journal of Accounting Research 44 (5): 9651000.
Krizan, Z., and P. D. Windschitl. 2007. The influence of outcome desirability on optimism.
Psychological Bulletin 133 (1): 95121.
Kunda, Z. 1990. The case for motivated reasoning. Psychological Bulletin 108 (3): 480498.
Leary, M. R., and R.M. Kowalski. 1990. Impression management: A literature review and two-
component model. Psychological Bulletin 107(1): 34-47

31
Lerner, J. S., and P. E. Tetlock. 1999. Accounting for the effects of accountability. Psychological
Bulletin 125 (2): 255275.
Lin, H., M. McNichols, and P. O'Brien. 2005. Analyst impartiality and investment banking
relationships. Journal of Accounting Research 43 (4): 623650.
McNichols, M., and P. OBrien. 1997. Self selection and analyst coverage. Journal of Accounting
Research 35 (Supplement): 167199.
Miller, J. S., and L. M. Sedor. 2014. Do stock prices influence analysts' earnings forecasts?
Behavioral Research in Accounting 26 (1): 85108.
Rosnow, R. L., and R. Rosenthal. 1995. "Some things you learn aren't so": Cohen's paradox,
Asch's paradigm, and the interpretation of interaction. Psychological Science 6 (I): 39.
Schlenker, B. R., and M. F. Weigold. 1992. Interpersonal processes involving impression
regulation and management. Annual Review of Psychology 43: 133-168.
Schoorman, F. D. and P. J. Holahan. 1996. Psychological antecedents of escalation behavior:
Effects of choice, responsibility, and decision consequences. Journal of Applied
Psychology 81 (6): 786-793.
Sedor, L. M. 2002. An explanation for unintentional optimism in analysts earnings forecasts.
The Accounting Review 77 (4): 731753.
Seybert, N. 2010. RandD capitalization and reputation-driven real earnings management. The
Accounting Review 85 (2): 671693.
Staw, B. M. 1981. The escalation of commitment to a course of action. Academy of Management
Review 6: 577-587.
Tetlock, P. E., L. Skitka, and R. Boettger. 1989. Social and cognitive strategies for coping with
accountability: Conformity, complexity, and bolstering. Journal of Personality and Social
Psychology 57 (4): 632640.
Wilks, T. J. 2002. Predecisional distortion of evidence as a consequence of real-time audit
review. The Accounting Review 77 (1): 5171.
Willis, R. H. 2001. Mutual fund manager forecasting behavior. Journal of Accounting Research
39 (3): 707725.

32
FIGURE 1
Ordinal Interaction between Holding and Dissemination Predicted by Hypothesis 1

Earnings
Forecast
interact
Public
dissemination

Private dissemination

No Holding Holding

33
FIGURE 2
Results for Hypothesis 1

Earnings
Forecast
20.5

20 Public
dissemination

Private
19.5 dissemination

19

18.5
No Holding Holding

34
TABLE 1
Overall Descriptive Statistics and ANOVA Results

Panel A: Mean earnings forecast (standard deviation)a


Holding

Dissemination Holding No Holding Overall


Public 20.38 19.17 19.81
(3.44) (2.70) (3.14)
n=39 n=35 n=74

Private 19.08 19.56 19.35


(3.24) (3.19) (3.19)
n=26 n=36 n=62

Overall 19.72 19.47 19.60


(3.35) (3.08) (3.16)
n=65 n=71 n=136

Panel B: ANOVA results


Source df Mean Square F p-value
Intercept 1 50754.30 5118.33 0.000
Holding 1 4.4801041 0.452 0.503
Dissemination 1 7.081 0.714 0.400
Holding x
1 23.763 2.396 0.062*
Dissemination

Panel C: Contrast-coded analysis of varianceb


Source SS df MS F p-value
Hypothesized Contrast 34.468 1 34.468 3.476 0.032*
Residual 0.856 2 0.428 0.043 0.957
Error 1308.937 132 9.916

Panel D: Test of simple effects


df t p-value
Holding > No Holding within Public 132 1.699 0.046*
Holding = No Holding within Private 132 0.606 0.545
Public > Private within Holding 132 1.684 0.047*
Public = Private within No Holding 132 0.528 0.599

35
a
Panel A reports mean earnings forecast and standard deviation of participants in each of the four
conditions. In the Holding condition, participants were told that the asset management firm had invested a
significant portion of its funds in a target stock based on the participants recommendation. In the No
Holding condition, they were told that the firm had not made any fund allocation to the stock. In the
Public condition, they were told that their forecast would be released to others, including colleagues and
clients. In the Private condition, they were told that their forecast would be kept private for their own use.
b
Panel C reports the results of a contrast-coded ANOVA with contrast weights of 3 for the
Public/Holding condition and -1 for the other conditions.

*one-tailed

36